【摘要】 为更准确地预测应急需求,以地震伤亡人数预测为例,收集1990—2010年全国范围内破坏性地震的伤亡人数资料,综合考虑影响地震人员伤亡的众多因素,选取地震发生时间、震级、震中烈度、人口密度、抗震设防烈度、预报水平等6项主要因素作为评价指标,采用BP神经网络,以主成分作为输入层神经元,伤亡人数作为输出层神经元,经过网络训练对样本数据进行仿真,建立地震伤亡人数预测模型。算例表明,与高斯拟合函数模型相比,BP神经网络模型对地震后伤亡人数的预测精度提高了7.5%。更多还原

【Abstract】 Earthquake casualty forecasting was improved with the model of BP neural network analysis.The data on earthquake casualties in China from 1990 to 2010 were collected as samples.Based on these data samples,some factors are selected as the main indicators,such as time of earthquake,earthquake magnitude,epicentral intensity,population density,resistance level and prediction level.First using the method of PCA to identify the principal components,then applying the model of BP neural network analysis to forecast the earthquake casualty.Finally,the proposed model and another forecasting model are used to predict the casualties in Yiliang earthquake(2012).The obtained result illustrates that the prediction accuracy is improved by 7.5%.更多还原